Summary: | The Arctic is considered the most susceptible environment to climate change, and as such has been of particular interest to climate change researchers. One such effect of climate change in the arctic is the increased incidence of thermokarst activity. Remote sensing technologies have been utilized to help detect thermokarst activity in the region, though typically thermokarst studies are multi-temporal analyses requiring multiple images and considerable time. This research sought to create a program that could detect thermokarst change from two images of different time periods. 1978 false color infrared aerial photographs, a 2005 SPOT image, and two Landsat images (2002 and 2006) encompassing the 20 kilometer radius around the Toolik Lake Field Station, Alaska, were utilized for this research. Spectral ranges were determined for each image and used as the criteria to numerically classify the two image pixels as water (1), thermokarst (3), and water (6). These images were then subtracted from each other to yield a numerical output containing information on the type of change, if any, that occurred on the landscape. These steps were compiled into one model for each possible change detection using Spatial Modeling Language (SML), and scripts were generated. ERDAS Macro Language (EML) was then used to create the graphical user interface that would allow the models to run based on user-input of before and after images and the creation of an output image. Program trials on three watersheds with known thermokarst activity and three without known thermokarst activity indicate that the program achieved its objective of identifying thermokarst activity, with an overall accuracy of 52%. The program was also able to identify areas that did not have any thermokarst activity, with an accuracy of 93.8%. The program did, however, identify lake perimeters as water change, whether as an increase or a decrease.
|